Limited "generics" in C without macros or UB June 5, 2017 on Drew DeVault's blog

I should start this post off by clarifying that what I have to show you today is not, in fact, generics. However, it’s useful in some situations to solve the same problems that generics might. This is a pattern I’ve started using to reduce the number of void* pointers floating around in my code: multiple definitions of a struct.

Errata: we rolled this approach back in wlroots because it causes problems with LTO. I no longer recommend it.

Let’s take a look at a specific example. In wlroots, wlr_output is a generic type that can be implemented by any number of backends, like DRM (direct rendering manager), wayland windows, X11 windows, RDP outputs, etc. The wlr/types.h header includes this structure:

struct wlr_output_impl;
struct wlr_output_state;

struct wlr_output {
    const struct wlr_output_impl *impl;
    struct wlr_output_state *state;
    // [...]

void wlr_output_enable(struct wlr_output *output, bool enable);
bool wlr_output_set_mode(struct wlr_output *output,
    struct wlr_output_mode *mode);
void wlr_output_destroy(struct wlr_output *output);

wlr_output_impl is defined elsewhere:

struct wlr_output_impl {
    void (*enable)(struct wlr_output_state *state, bool enable);
    bool (*set_mode)(struct wlr_output_state *state,
        struct wlr_output_mode *mode);
    void (*destroy)(struct wlr_output_state *state);

struct wlr_output *wlr_output_create(struct wlr_output_impl *impl,
        struct wlr_output_state *state);
void wlr_output_free(struct wlr_output *output);

Nowhere, however, is wlr_output_state defined. It’s left an incomplete type throughout all of the common wlr_output code. The “generic” part is that each output implementation, in its own private headers, defines the wlr_output_state struct for itself, like the DRM backend:

struct wlr_output_state {
    uint32_t connector;
    char name[16];
    uint32_t crtc;
    drmModeCrtc *old_crtc;
    struct wlr_drm_renderer *renderer;
    struct gbm_surface *gbm;
    EGLSurface *egl;
    bool pageflip_pending;
    enum wlr_drm_output_state state;
    // [...]

This allows implementations of the enable, set_mode, and destroy functions to avoid casting a void* to the appropriate type:

static struct wlr_output_impl output_impl = {
    .enable = wlr_drm_output_enable,
    // [...]

static void wlr_drm_output_enable(struct wlr_output_state *output,
        bool enable) {
    struct wlr_backend_state *state =
        wl_container_of(output->renderer, state, renderer);
    if (output->state != DRM_OUTPUT_CONNECTED) {
    if (enable) {
        // [...]
    } else {

// [...]
struct wlr_output output = wlr_output_create(&output_impl, output);

The limitations of this approach are apparent: you cannot work with multiple definitions of wlr_output_state in the same file. However, you get improved type safety, have to write less code, and improve readability.

Articles from blogs I read Generated by openring

Status update, April 2024

Hi! The X.Org Foundation results are in, and I’m now officially part of the Board of Directors. I hope I can be of use to the community on more organizational issues! Speaking of which, I’ve spent quite a bit of time dealing with Code of Conduct matters latel…

via emersion April 16, 2024

M2dir: treating mails as files without going crazy

Sometime recently in the past I complained about Maildir. You can go read the post, but the executive summary is that I think Maildir uses an actively user-hostile directory structure and extremely convoluted filenames that do not convey any meaning at all. …

via blogfehler! April 15, 2024

Go Developer Survey 2024 H1 Results

What we learned from our 2024 H1 developer survey

via The Go Blog April 9, 2024